Study on multi-view calibration method in non-contact measurement

Abstract

This study do binocular calibration experiments on the basis of the Z. Zhang calibration method, then unify coordinate system which can determine the position of the camera with respect to the world. The whole process use image processing software halcon to perform the multi-view calibration. It is argued that the method can rapidly, accurately achieve the calibration of eight cameras in convergence mode. Introduction In non-contact measurement the measurement technology using image measuring technique can overcome some problems in the traditional measuring method. It can achieve high efficiency, automation, dynamic testing requirements. Therefore, applying image measuring technique to geometric measurement is one of the important development direction of geometric measurement technology in the future. The calibration of image measurement system is the basis and prerequisite of acquiring 3D information adopting computer vision technology and is also a key link to improve 3D measurement accuracy [4]. It is to build the corresponding points relationship between space object 3D world coordinates and the 2D image coordinates [3]. By the relationship we can recover the depth information from dealing with collected 2D images. So camera calibration is a very important step. Because the measurement filed is large in experiments, the monocular or binocular doesn’t realize measurement requirement. In order to meet the requirement, we use eight cameras. For now, binocular calibration method is mature, for example, Tsai radial alignment constraint two-step method [1], Z. Zhang plane pattern two-step method, etc. Z. Zhang calibration method [2] [5] is applied in the experiment to get two adjacent camera’s position. This part of the principle is no longer described in this paper [2] [5]. Binocular vision system in convergence mode Binocular vision system is the simulation of human vision. It generally uses the relative position of two cameras to shoot the same object. Two or more images need to be collected. By these images compute the parallax of same points in the scenecaptured in two cameras. Then use triangulation method to get the space point coordinates. Fig. 1 Parallax figure of binocular imaging in convergence mode International Conference on Advances in Mechanical Engineering and Industrial Informatics (AMEII 2015) © 2015. The authors Published by Atlantis Press 509 In this work, optical axis of two cameras is not parallel but in a certain angle around the centers of their projection. Namely two cameras are in convergence mod, as show in Fig. 1. The baseline is located in x-z plane. With a space point Pw corresponding to image point(x1,y1)and (x2,y2) respectively known, the optical axes of two cameras intersect at a2α angle to produce a point K or (0,0,z) in the X-Z plane. Make two perpendicular to z1 and z2 axes respectively through the point Pw, and we can know that the angles between the two perpendiculars and the X axis are both α. By use of similar triangle calculations: